resnet-image-embedding/pytorch/model.py

# Copyright 2021 Zilliz. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.


import torch
from torch.nn import Linear
from torch import nn
import timm


class Model():
    """
    PyTorch model class
    """

    def __init__(self, model_name, num_classes=1000):
        super().__init__()
        self._model = timm.create_model(model_name, pretrained=True)
        pretrained_dict = None
        if model_name == 'resnet101':
            pretrained_dict = torch.hub.load_state_dict_from_url(
                'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-rsb-weights/resnet101_a1h-36d3f2aa.pth')
        if model_name == 'resnet50':
            pretrained_dict = torch.hub.load_state_dict_from_url(
                'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-rsb-weights/resnet50_a1_0-14fe96d1.pth')
        if pretrained_dict:
            self._model.load_state_dict(pretrained_dict, strict=False)
        if num_classes != 1000:
            self.create_classifier(num_classes=num_classes)
        self._model.eval()

    def __call__(self, img_tensor: torch.Tensor):
        self._model.eval()
        features = self._model.forward_features(img_tensor)
        if features.dim() == 4:  # if the shape of feature map is [N, C, H, W], where H > 1 and W > 1
            global_pool = nn.AdaptiveAvgPool2d(1)
            features = global_pool(features)
        return features.flatten().detach().numpy()

    def create_classifier(self, num_classes):
        self._model.fc = Linear(self._model.fc.in_features, num_classes, bias=True)